HCIN 620 Lab 6 Course Project

In this project we are tasked with predicting the stages of Chronic Kidney Disease based on Glomerular Filtration Rate (GFR). Information for
dataset is available in this link http://archive.ics.uci.edu/ml//datasets/Chronic_Kidney_Disease

In order to succeed in this nal you will be required to input most of the python code. Please review all the previous labs before you begin, and
read the instructions carefully.

Rather than using a question/answer format, we have commented the code cells with the notation #TODO This is a placeholder technique (a To
Do list, so to speak) commonly used in machine learning. Complete each #TODO task requested of you.

Good Luck!

Notebook by Reza Afra, Ph.D. and Barbara Berkovich, Ph.D., M.A.

Last update December 28, 2020

Step 1: Environment Setup

You have learned that in order to setup an environment for your project, you'd need to rst import the libraries you need.

#ANSWER KEY TODO 1. Import the libraries all the libraries

#Extra libraries are okay, and they only need to have the minimum to complete
#this project without errors.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
plt.style.use('ggplot')

from sklearn.preprocessing import OneHotEncoder, StandardScaler

from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.metrics import r2_score, mean_squared_error, accuracy_score, confusion_matrix
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.compose import ColumnTransformer

# Suppress pesky warnings

import warnings
warnings.filterwarnings("ignore")

#ANSWER KEY TODO 2. Add a print command to acknowledge completion of the import.

print('Import complete')

Import complete

Step 2: Data Cleaning

Upload the data le called data-lab-6-ckd-courseproject.csv data.

Read the csv le into a data frame, and use the name of the dataframe to print the rst and last 5 rows.

#ANSWER KEY TODO 3. Load the data into a pandas dataframe

df = pd.read_csv('data-lab-6-ckd-courseproject.csv')
data = df.copy()

#ANSWER KEY TODO 4. Use the name of the dataframe to print the first and last five rows.

data

Red Blood
Blood Specific Pus Pus Cell Blood Serum
Age Albumin Sugar Blood Bacteria Glucose Sodium P
Pressure Gravity Cell clumps Urea Creatinine
Cells Random

0 48.0 70.0 1.005 4.0 0.0 normal abnormal present notpresent 117.0 56.0 3.8 111.0

1 53.0 90.0 1.020 2.0 0.0 abnormal abnormal present notpresent 70.0 107.0 7.2 114.0

2 63.0 70.0 1.010 3.0 0.0 abnormal abnormal present notpresent 380.0 60.0 2.7 131.0

3 68.0 80.0 1.010 3.0 2.0 normal abnormal present present 157.0 90.0 4.1 130.0

4 61.0 80.0 1.015 2.0 0.0 abnormal abnormal notpresent notpresent 173.0 148.0 3.9 135.0

... ... ... ... ... ... ... ... ... ... ... ... ... ...

153 55.0 80.0 1.020 0.0 0.0 normal normal notpresent notpresent 140.0 49.0 0.5 150.0

154 42.0 70.0 1.025 0.0 0.0 normal normal notpresent notpresent 75.0 31.0 1.2 141.0

155 12.0 80.0 1.020 0.0 0.0 normal normal notpresent notpresent 100.0 26.0 0.6 137.0

156 17.0 60.0 1.025 0.0 0.0 normal normal notpresent notpresent 114.0 50.0 1.0 135.0

157 58.0 80.0 1.025 0.0 0.0 normal normal notpresent notpresent 131.0 18.0 1.1 141.0

158 rows × 25 columns

data.isnull().sum()

Age 0
Blood Pressure 0
Specific Gravity 0
Albumin 0
Sugar 0
Red Blood Cells 0
Pus Cell 0
Pus Cell clumps 0
Bacteria 0
Blood Glucose Random 0
Blood Urea 0
Serum Creatinine 0
Sodium 0
Potassium 0
Hemoglobin 0
Packed Cell Volume 0
White Blood Cell Count 0
 Red Blood Cell Count 0
Hypertension 0
Diabetes Mellitus 0
Coronary Artery Disease 0
Appetite 0
Pedal Edema 0
Anemia 0
Class 0
dtype: int64

le=LabelEncoder()
data.drop_duplicates()

Red Blood
Blood Specific Pus Pus Cell Blood Serum
Age Albumin Sugar Blood Bacteria Glucose Sodium P
Pressure Gravity Cell clumps Urea Creatinine
Cells Random

0 48.0 70.0 1.005 4.0 0.0 normal abnormal present notpresent 117.0 56.0 3.8 111.0

1 53.0 90.0 1.020 2.0 0.0 abnormal abnormal present notpresent 70.0 107.0 7.2 114.0

2 63.0 70.0 1.010 3.0 0.0 abnormal abnormal present notpresent 380.0 60.0 2.7 131.0

3 68.0 80.0 1.010 3.0 2.0 normal abnormal present present 157.0 90.0 4.1 130.0

4 61.0 80.0 1.015 2.0 0.0 abnormal abnormal notpresent notpresent 173.0 148.0 3.9 135.0

... ... ... ... ... ... ... ... ... ... ... ... ... ...

153 55.0 80.0 1.020 0.0 0.0 normal normal notpresent notpresent 140.0 49.0 0.5 150.0

154 42.0 70.0 1.025 0.0 0.0 normal normal notpresent notpresent 75.0 31.0 1.2 141.0

155 12.0 80.0 1.020 0.0 0.0 normal normal notpresent notpresent 100.0 26.0 0.6 137.0

156 17.0 60.0 1.025 0.0 0.0 normal normal notpresent notpresent 114.0 50.0 1.0 135.0

157 58.0 80.0 1.025 0.0 0.0 normal normal notpresent notpresent 131.0 18.0 1.1 141.0

158 rows × 25 columns

The data is located at 'https://drive.google.com/ le/d/19oQaKN4NQiGa6Tq9P18Zk3ImbKgzIw9m/view?usp=sharing' . In the following cell,

connect to the data and print a statement that indicates you've successfully completed this step.

# Load the data into a pandas dataframe

url = 'https://drive.google.com/file/d/19oQaKN4NQiGa6Tq9P18Zk3ImbKgzIw9m/view?usp=sharing'
path = 'https://drive.google.com/uc?export=download&id='+url.split('/')[-2]
df = pd.read_csv(path)
data = df.copy()
print('Data connection complete')

Data connection complete

Add more data cleaning...

Step 3: Exploratory Data Analysis (EDA) and Preprocessing

Next we're going to build the targets which are stages of CKD in a column we will call "CKD Stages" There are various equations for calculating
GFR but here we will stick with a simpli ed form of it. Please read about GFR in the following link. Source: https://www.niddk.nih.gov/health-
information/professionals/clinical-tools-patient-management/kidney-disease/laboratory-evaluation/glomerular- ltration-rate/estimating

# Used a formula given by NIDDK which was simpler and made it even simpler by
# removing the effect of race and gender
# GFR (mL/min/1.73 m^2) = 175 × (Scr)^-1.154 × (Age)-0.203

# DO NOT change the code below

def calc_gfr(Scr, Age):
return (175 * (Scr) ** -1.154) * (Age ** -0.203)

# I tried 6 stages but the dataset is too small so classes 4 and 5

# had no instances. I reduced the number of classes to 3.
def calc_ckd_stage(gfr):
bins = [0, 45, 90, 250]
labels = [3, 2, 1]
ret = pd.cut(gfr, bins=bins, labels=labels)
return ret

# DO NOT change the code below

data["GFR"] = calc_gfr(data["Serum Creatinine"], data["Age"])
gfr = data['GFR']
removed_outliers = gfr.between(gfr.quantile(.00), gfr.quantile(.95))

data = data[removed_outliers]

# DO NOT change the code below

data["CKD Stages"] =calc_ckd_stage(data["GFR"])

# DO NOT change the code below

data["CKD Stages"].value_counts()

2 53
1 45
3 39
Name: CKD Stages, dtype: int64

Histogram

# TODO: Print a histogram of the values of "Serum Creatinine"

data["Serum Creatinine"].hist();
Scatterplot

# TODO: Print a scatterplot of the values of "GFR" vs "Serum Creatinine"

sns.scatterplot(data["Serum Creatinine"], data["GFR"]);

Isolate features from target

# TO DO: Isolate features and targets

X = data.drop(['Class', 'CKD Stages'], axis=1)

y = data["CKD Stages"]

There are various ways to encode categorical data. You learned about some of them during labs. Visit this link and read it thoroughly. Find an
appropriate encoding scheme and transform the categorical attributes of your dataset.

# TO DO: Isolate the categorical features and encode them using a proper encoding method.
# TO DO: Isolate the numverical features and scale them using a proper scaling method.

numerical_ix = X.select_dtypes(include=['float64']).columns
categorical_ix = X.select_dtypes(include=['object']).columns

#define the data preparation for the columns

t = [('cat', OneHotEncoder(), categorical_ix), ('num', StandardScaler(), numerical_ix)]
transform = ColumnTransformer(transformers=t)

X = transform.fit_transform(X)

#TO DO: Print a heatmap of correlation between features

plt.figure(figsize=(8, 8))

sns.heatmap(data.corr(), cbar=True, annot=False, yticklabels=numerical_ix,

xticklabels=numerical_ix);

Split the Data

# TODO: Split the data to train and test sets

X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=data['CKD Stages'], random_state=308)

Step 4: Build the Models and Evaluate

Logistic Regression
# Use Logistic Regression to predict the stage of the kidney function
# and print the accuracies of your model on both train and test data
log_reg = LogisticRegression()
log_reg.fit(X_train, y_train)
y_pred = log_reg.predict(X_test)
print(f' Accuracy on test set: {accuracy_score(y_test, y_pred):.3f}')
print(f' \nAccuracy on train set: {accuracy_score(y_train, log_reg.predict(X_train)):.3f}')

Accuracy on test set: 0.971

Accuracy on train set: 0.990

Confusion Matrix

# TODO: Plot a Confusion Matrix based on your findings from the previous step

# TODO: Plot a Confusion Matrix based on your findings from the previous step

data_ = {'y_true': y_test,

'y_pred': y_pred
}

df = pd.DataFrame(data_, columns=['y_true','y_pred'])
confusion_matrix = pd.crosstab(df['y_true'], df['y_pred'], rownames=['ACTUAL'], colnames=['PREDICTED'])

sns.heatmap(confusion_matrix, annot=True)
plt.show()

K-nearest Neighbors

# TO DO: Use Logistic Regression to predict the stage of the kidney function
# and print the accuracies of your model on both train and test data
knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(X_train, y_train)

y_pred = knn.predict(X_test)
print(f' Accuracy on test set: {accuracy_score(y_test, y_pred):.3f}')

Accuracy on test set: 0.857

# TO DO: Find the optimal number of neighbors and print the accuracies
# on both test and train data for that number of neighbors.
accuracies = []
for N in range(1,20):
knn = KNeighborsClassifier(n_neighbors=N)
knn.fit(X_train, y_train)
y_pred = knn.predict(X_test)
acc = accuracy_score(y_test, y_pred)
accuracies.append(acc)

accuracies = np.array(accuracies) # convert to numpy array

sns.lineplot(np.arange(1,20),accuracies);
# Find the best k
best_k = 1 + np.argmax(accuracies) # add one b/c arrays are 0-indexed
best_accuracy = np.max(accuracies)
print(f"Best k: {best_k} \nBest Accuracy from kNN: {best_accuracy:.3f}")

Best k: 4
Best Accuracy from kNN: 0.886

If your code runs cleanly all the way through, then print as a pdf and submit to Blackboard Module 6 for grading.
 0s completed at 11:06 PM